Using it with newer whisper models like whisper-large-v3?

[empz]

Is this library compatible with newer Whisper models (v2, v3)?

I've tried using the whisper-large-v3, but transcribe throws the following error:

RuntimeError: Given groups=1, weight of size [1280, 128, 3], expected input[1, 80, 3000] to have 128 channels, but got 80 channels instead

[Jeronymous]

It should work with large V3
What is your version of whisper-timespamped? And openai-whisper.
If it's not the latest version for both, updating should fix.

If if fails with the latest version, please provide a minimal code to reproduce.

[empz]

Uhm, it's working now. No idea why though. I've just tried this for the first time yesterday in 2 different platforms (Google Colab and Modal) and it was failing. I guess an older version of some package was being installed for some reason...